Table 1.

Prediction accuracy comparison between pathologists and AI on top-100 loss images. p-values were obtained from the McNemar test.

	Forensic pathologists		AI		p-value
Gunshot wound	Correct	Incorrect	Correct	Incorrect
Entrance (n = 54)	43 (80%)	11 (20%)	5 (9%)	49 (91%)	<.0001*
Exit (n = 46)	13 (28%)	33 (72%)	10 (22%)	36 (78%)	.58
Total (n = 100)	56 (56%)	44 (44%)	15 (15%)	85 (85%)	<.0001*

Entrance GSW		AI
		Correct	Incorrect
Pathologists	Correct	4	39
	Incorrect	1	10
Exit GSW		AI
		Correct	Incorrect
Pathologists	Correct	5	8
	Incorrect	5	28
Total		AI
		Correct	Incorrect
Pathologists	Correct	9	47
	Incorrect	6	38